Optimizing Energy Consumption in Electric Vehicles: A Systematic and Bibliometric Review of Recent Advances

Abstract

Electric vehicles are key to sustainable mobility, but their limited range remains a major obstacle to widespread adoption. Extending driving distance requires optimizing energy use across subsystems. This study combines bibliometric mapping (2017–2024, Scopus) with a focused qualitative review to structure recent research. Results highlight a strong emphasis on energy efficiency, with China leading due to its market size, industrial base, and supportive policies. Major research directions tied to range extension include energy storage, motion control, thermal regulation, cooperative driving, and grid interaction. Among these, hybrid energy storage systems and motor control stand out for their measurable impact and industrial relevance, while thermal management, regenerative braking, and systemic approaches (V2V and V2G) remain underexplored. Beyond mapping contributions, the study identifies ongoing gaps and calls for integrated strategies that combine electrical, thermal, and mechanical aspects. As EV adoption accelerates and battery demand increases, the findings emphasize the need for battery-aware, multi-objective energy management strategies. This synthesis provides a vital framework to guide future research and support the development of robust, integrated, and industry-ready solutions for optimizing EV energy use and extending driving range.

1. Introduction

In response to the environmental emergency, governments and regulatory bodies are enforcing increasingly stringent policies to reduce pollutant emissions and promote environmentally friendly practices. Among the sectors with the highest environmental footprint, road transport stands out due to its substantial share of global CO₂ emissions, based on typical driving cycles [1]. Despite certain improvements, emissions from this sector have continued to rise, further intensifying their impact on the climate. This upward trend is particularly alarming, given that some segments of road transport remain especially energy-intensive and polluting. Diesel-powered vehicles, in particular, account for a significant share—currently close to a quarter of the sector’s total emissions. Without decisive intervention, this figure is expected to considerably increase by 2030 [2]. Their growing presence on the roads and their frequent use only worsen the situation, making it essential to adopt alternative solutions capable of reducing their environmental footprint [3].

In this context, clean propulsion technologies, such as electric vehicles, are emerging as an essential solution. By significantly reducing emissions and improving air quality [4,5], these technologies align closely with the recommendations made to policymakers and stakeholders within the framework of the United Nations Agenda 2030 for Sustainable Development Goals [6]. By rethinking mobility from a more sustainable perspective, these innovations pave the way for a future in which road transport can finally meet current environmental standards.

Despite their apparent advantages, electric vehicles (EVs) still face significant challenges [7] that hinder their large-scale adoption. The main obstacle remains the current limitation of driving range, which continues to fuel drivers’ anxiety about reaching their destinations on a single charge. This “range anxiety” is a significant barrier to broader acceptance, especially for long journeys or in regions where charging infrastructure is scarce [8,9]. To overcome these concerns, significant improvements in the management and optimization of energy consumption are needed. The main goal is to increase the distance traveled per charge while maintaining high performance and ensuring battery longevity. In this context, the growing interest in energy consumption optimization strategies stems from their potential to improve overall efficiency. Better energy efficiency not only enhances the practicality of EVs but also strengthens their competitive advantage in the transport market.

Recent research on electric vehicles shows that improving driving range and optimizing energy consumption does not rely on a single approach or a unique pathway, but rather on a variety of strategies explored across the literature. Each study tends to adopt a specific perspective, focusing on a particular subsystem of the vehicle. Some works concentrate on the design and optimization of the powertrain [10,11,12], aiming to enhance overall efficiency and reduce energy losses by refining key components such as the motor, battery, converters, gearbox, and transmission. Other studies focus on the management of auxiliary systems [13,14,15,16,17], seeking to reduce the energy demand of onboard equipment (air conditioning, heating, lighting, and electronic devices) while maintaining comfort and safety. A further line of research emphasizes control and energy management strategies [18,19,20], which aim to allocate the available energy intelligently according to driving conditions, state of charge, and power demand. Each of these directions provides specific benefits but also carries inherent limitations.

The first category concentrates on the vehicle’s hardware and energy setup. Increasing battery capacity [21] can indeed extend driving range. However, this method has several structural drawbacks: vehicle overweight, long charging times, and high initial and environmental costs [22]. Additionally, it does not address long-term sustainability issues caused by reliance on critical materials such as lithium, cobalt, and nickel. Other studies suggest a more integrated approach to hardware sizing. Huertas and colleagues [10] provide a relevant example: by optimizing motor size, battery capacity, and transmission ratios together for a specific use, they show that significant improvements in performance and efficiency are possible. Their findings notably indicate a 37% reduction in energy consumption compared to an equivalent diesel vehicle, while also improving acceleration and lowering costs and emissions. This work demonstrates that optimal sizing of powertrain hardware components can be a more effective alternative than simply increasing onboard energy capacity. However, this approach also has limitations: it works best in well-defined contexts (such as short, regular trips) and does not guarantee the same benefits for long-distance, heavy freight, or services with high route variability.

A significant portion of EV research does not focus on increasing the amount of available energy but rather on optimizing its use. In this regard, Energy Management Strategies (EMSs) constitute a central focus [21]. The literature mainly distinguishes two major families. The first group comprises rule-based approaches [20,21,22]. Simple to implement and intuitive, they adjust power distribution according to vehicle status and driving conditions, whether using deterministic rules or more advanced variants such as fuzzy logic [22,23,24,25,26,27]. Their main advantage lies in their low computational complexity, making them easily deployable on board. However, their rigidity limits their effectiveness in complex and dynamic environments: these methods struggle to adapt to variable traffic conditions or changing road profiles. The second family corresponds to optimization-based approaches [28,29,30,31,32,33], which rely on mathematical models and cost functions to determine the most efficient real-time power distribution. These methods use a systemic representation of the vehicle and its environment, allowing them to manage several objectives (reducing consumption, minimizing losses, and extending battery life) and constraints (thermal limits and component capacities) simultaneously. These approaches fall into two subcategories. Global optimization seeks to reduce consumption over an entire predetermined driving cycle, using techniques such as dynamic programming [28,29], genetic algorithms [34,35,36], etc. Although effective in providing reference solutions, this method is not well-suited to real-time use due to its computational complexity and the need for complete driving profile data. Instantaneous optimization, on the other hand, aims to minimize consumption at each instant of the driving cycle. Strategies such as Model Predictive Control (MPC) and Equivalent Consumption Minimization Strategy (ECMS) are emblematic examples. MPC anticipates future system behavior using a dynamic model and proactively adjusts commands, enabling continuous fine optimization directly translated into control actions [37,38,39]. ECMS, meanwhile, transforms the complex dynamic optimization problem into a series of instantaneous problems, seeking to minimize equivalent fuel (or energy) consumption at each time step [40]. These methods are particularly effective for embedded control as they combine theoretical rigor with practical feasibility.

According to the literature, these strategies (rule-based or optimization-based) have demonstrated some effectiveness, but they also present significant limitations. They remain largely reactive, relying mainly on internal variables such as battery state of charge, vehicle speed, or instantaneous power demand. In other words, they only respond to immediate needs without real anticipation capability. Moreover, these approaches rarely consider external factors such as traffic density, road topography, or weather conditions, even though these parameters strongly influence energy consumption and overall EV range [41].

Recently, research has shown that artificial intelligence (AI) [42,43,44] and, in particular, reinforcement learning (RL) and its deep variants (Deep Q-Learning and Deep Reinforcement Learning) [45,46,47,48] represent a breakthrough in EV energy management. According to Kumar and Kukker [49], integrating an RL-based EMS can reduce energy consumption by 15–20%. Similarly, another study [50] applied Deep Reinforcement Learning (DRL) and demonstrated that its achieved efficiency reached 87.2–90.7% of that of reference strategies based on dynamic programming, while reducing unnecessary engine operation cycles. More recently, an energy management strategy based on Double Deep Q-Learning (DDQL) was proposed for fuel cell hybrid electric vehicles [51]. The results indicate a 15.4% improvement compared to a rule-based strategy in training scenarios and 13.3% in test scenarios, confirming the potential of deep RL to optimize consumption while respecting the dynamic constraints specific to fuel cell systems. These contributions highlight the central role of AI in developing predictive and adaptive strategies capable of anticipating real-time energy needs and managing complex, partially observable, or non-stationary environments. However, the authors also point out several challenges, notably the availability and representativeness of training data, the explainability of algorithmic decisions, and robustness in unforeseen scenarios [52].

The examination of this literature reveals a major observation: existing work remains fragmented and highly specialized. Each study addresses energy optimization from a specific angle, whether it involves the powertrain, auxiliary systems, or energy management strategies, using distinct methods. Although relevant, these contributions remain compartmentalized and do not offer an integrated vision enabling a systematic comparison of approaches according to the vehicle dimensions considered. This fragmentation makes it challenging to identify coherent paths toward innovation and to highlight dominant trends and actual research gaps. This lack of structuring fully justifies the use of a bibliometric analysis approach, which is capable of organizing and mapping the directions pursued by the scientific community.

Bibliometrics is a field of research that provides quantitative and objective tools for analyzing scientific production, knowledge networks, and academic impact. It is based on the systematic study of publications and citations, and thus provides a solid basis for assessing the dynamics and evolution of science [53]. Beyond its descriptive role, it also focuses on building and interpreting indicators to measure the influence of studies and guide research policies [54]. By relying on mathematical and statistical methods, it models the processes of scientific publishing and communication and highlights relationships between researchers, institutions, and disciplines [55]. More recently, bibliometric analysis has been conceived as a systematic and operational process, integrating well-defined steps from data collection and cleaning to visualization and interpretation of results [56]. Bibliometrics and bibliometric analysis are now positioned as essential tools for understanding research trends, identifying emerging fields, and evaluating the impact of scientific contributions.

Several review studies have used bibliometric analysis to explore the field of EVs, but their objectives and scopes differ significantly. Some aim to characterize the overall scientific rise of EVs. This is the case of Barbosa et al. [57], who adopted a broad approach centered on the generic term “electric vehicle” to provide an overview of emerging technological themes, from propulsion innovations to infrastructure developments. Similarly, Ullah et al. [58] analyzed publications from 2011 to 2022 using co-occurrence, citation, and co-author mapping to trace the progressive structuring of the field.

Other studies have focused on specific components of the EV ecosystem. Veza et al. [59], for example, examined lithium-ion batteries, battery management systems (BMSs), charging infrastructures, and vehicle-to-everything (V2X) interactions, identifying recent research trends in these subdomains. Meanwhile, Miah et al. [60] conducted a bibliometric analysis specifically focused on energy management strategies (EMSs) between 2010 and 2021, emphasizing optimization methods developed for EVs. On a more socio-technical level, Bhat et al. [61] analyzed behavioral and structural factors influencing EV adoption in developing countries. None of these analyses, however, offered a truly integrated vision of optimizing energy consumption and range in EVs.

A bibliometric review of a corpus of recent articles (2017–2024) makes it possible to organize existing contributions, map dominant research directions, and reveal ongoing underexplored themes on EV energy optimization. By combining this quantitative approach with a targeted qualitative review, the present work proposes the first structured mapping of scientific directions in this field, identifying major dynamics, persistent bottlenecks, and future innovation opportunities.

The remainder of this article is structured as follows:

Section 2 details the methodological approach used for the bibliometric analysis. It outlines the process of data retrieval from the Scopus database, specifying the search queries and selection criteria. The section also describes the analytical tools and techniques employed to examine and interpret the scientific landscape. In Section 3, the findings are presented and discussed. This includes an initial bibliometric overview of global research trends, followed by a thematic analysis of the key scientific directions related to energy optimization in electric vehicles. Lastly, Section 4 concludes the article by summarizing the main contributions and proposing avenues for future research aimed at developing more effective and context-aware energy management strategies.

2. Materials and Methods

Methodological Approach: Bibliometric Analysis

Bibliometrics, first introduced by Paul Otlet in the 1930s, was later redefined by Pritchard in 1969 as the use of statistical methods to study scientific literature [62,63,64]. Since then, it has become a distinct field within information sciences, providing powerful tools to quantitatively analyze academic output based on published data and related metadata, such as titles, abstracts, authors, and keywords [65]. This method allows for organizing large amounts of information, identifying key themes, pinpointing the most influential contributions, and mapping collaboration networks. It also helps uncover underexplored research areas, spot emerging trends, and direct future research toward innovative and promising directions [66,67].

In this study, bibliometric analysis is employed to map the main research dynamics related to energy optimization in electric vehicles, to provide a clear, coherent, and structured synthesis of the current state of knowledge in this field.

The bibliometric analysis was designed following a rigorous methodology inspired by widely accepted standards in the literature. It is structured into several successive stages to extract, process, and explore relevant scientific data [56]. This approach involves selecting a specialized database, defining search criteria, cleaning the collected data, and analyzing it using bibliometric analysis and visualization tools. The overall methodological process is summarized in Figure 1, and each step is described in detail in the following sections.

The choice of a suitable database is an essential step in any bibliometric study. Some studies rely on several sources simultaneously, while others prefer a single database to guarantee metadata consistency and avoid redundancy.

In the context of this research, the selection of the database was guided by the reading of recent articles comparing the leading platforms dedicated to bibliographic research, according to various criteria such as disciplinary coverage, indexing quality, richness of metadata, and export functionalities [68]. These analyses highlighted the relevance of the Scopus database, regularly cited as one of the most reliable and comprehensive [69], particularly in the fields of engineering, energy, and intelligent systems. The quality of its indexing, the structure of its metadata, and the compatibility of its export system with bibliometric analysis tools make it a reference resource in scientific literature. Consequently, all the data used in this study were extracted exclusively from the Scopus database.

• Phase 1: Literature Search Strategy and Selection Criteria

The bibliographic data collection was conducted in 2024 using a structured query on the Scopus database. This process involved two primary stages: defining search terms and establishing inclusion and exclusion criteria. The search terms were designed to comprehensively cover all approaches potentially contributing to the optimization of energy consumption in electric vehicles. This included energy management techniques, control strategies, algorithms, and other methods directly related to energy performance.

Unlike studies that focus exclusively on conventional electric vehicles, this work aimed for a broader scope, encompassing all types of electric vehicles, including autonomous and intelligent variants. The objective was to provide a global perspective on recent contributions without excluding emerging innovations in the field.

To ensure the relevance and originality of the analysis, the bibliometric study focused on publications published between 2017 and 2024. This time frame was deliberately selected to build upon earlier bibliometric studies such as [60,70,71], which offered initial mappings of the field, but did not cover the recent surge of research efforts dedicated to energy optimization in electric vehicles. By concentrating on this updated period, the study aims to provide a more structured and contemporary view of emerging scientific trends, methodological developments, and newly established research directions that have gained momentum in recent years. No language filter was applied, ensuring the widest possible coverage and allowing relevant contributions to be captured regardless of their publication language. This decision prioritized the richness of the dataset over linguistic limitations. Additionally, the search was limited to scientific disciplines closely related to the topic of interest to exclude publications outside the relevant technical and scientific scope. Only peer-reviewed journal articles and literature reviews were included, deliberately excluding other types of documents such as conference proceedings, book chapters, notes, errata, editorials, short surveys, meeting abstracts, books, letters, retracted articles, data papers, and reports. This selection ensured methodological consistency and high academic rigor across the sources analyzed.

All filtering steps and the number of retained documents at each stage are presented in

Table 1. Overview of automated filtering steps performed in Scopus.

Filtering Step	Description	Search Query/Filter	Filtered Documents
Initial identification	Retrieval of all records related to energy optimization in electric vehicles	TITLE-ABS-KEY ((“energy optimization” OR “energy consumption reduction” OR “energy management”) AND (“electric vehicle *” OR “autonomous electric vehicle *”) AND (“strategy” OR “method *” OR “algorithm *”))	8365
Subject area filtering	Restriction to disciplines most relevant to the research	LIMIT-TO (SUBJAREA, “ENGI” OR “ENER” OR “COMP” OR “MATH” OR “ENVI”)	8162
Publication year filtering	Limitation to documents published between 2017 and 2024 to focus on recent developments	PUBYEAR > 2016 AND PUBYEAR < 2025	6138
Document type filtering	Selection of peer-reviewed articles and literature reviews only	LIMIT-TO (DOCTYPE, “ar” OR “re”)	3853

Note: Filtering steps follow standard Scopus-based bibliometric methodologies [56,72,73].

. (The asterisk (*) is used as a truncation symbol in Scopus to include all possible endings of a word).

• Phase 2: Final Selection of Relevant Articles

Following the initial extraction of results from the Scopus database, a large number of articles were retrieved. Although filters were applied within the database (thematic focus, publication period, document type), the resulting dataset still contained potentially irrelevant publications. Therefore, a second selection phase was necessary to ensure that only documents directly related to the research topic were retained. This screening step was carried out after exporting the metadata, using filtering and sorting tools. In the first stage, article titles and Digital Object Identifiers (DOIs) were used to detect and remove duplicates. This initial cleaning process helped consolidate a reliable working dataset. Next, keyword filtering was applied to the Title, Abstract, and Keywords columns using automated formulas. The goal was to identify publications focused on electric vehicles by detecting the presence of expressions such as “electric vehicle,” “autonomous vehicle,” “EV,” “AEV,” “self-driving vehicle,” “autonomous electric vehicle,” and similar terms. Each cell returned a “Yes” or “No” depending on the presence of these terms.

To avoid inadvertently excluding relevant research that may have used alternative or indirect terminology, a complementary manual review was conducted. After the automated filtering steps, a closer reading of titles and abstracts was performed to discard publications that were not closely aligned with the core issue of energy management in electric vehicles. Finally, a thorough manual review was carried out to confirm the exact relevance of each selected article, based on its scientific content and its contribution to the topic. At the end of this process, the dataset was refined to include only the most pertinent articles, forming the final corpus analyzed in this study. The different stages of this manual filtering process, applied after the initial Scopus data extraction, are summarized in

Table 2. Filtering Steps Applied After Data Extraction.

Filtering Step	Description	Purpose
Duplicate removal	Identification and elimination of duplicate records using titles and DOIs	To ensure no redundancies in the dataset
Automated keyword filtering	Use of spreadsheet functions to detect keywords in the title, Abstract, and Keywords columns related to electric vehicles	To exclude publications not explicitly referring to electric vehicles
Manual check of excluded items	Quick review of documents labeled “No” to recover potentially relevant articles that used alternative wording	To avoid unintentional omission of relevant studies
Screening of titles and abstracts	Manual review to exclude articles not addressing energy optimization in electric vehicles	To retain only studies with direct relevance to the research objective
Final full-text relevance check	In-depth examination of remaining articles to confirm their scientific relevance and alignment with the research scope	To establish a final, high-quality, and coherent article set for bibliometric analysis

Note: The filtering procedure was designed following standard practices in bibliometric research [56,72,73].

. The article selection process was conducted and reported in accordance with the PRISMA 2020 guidelines [74]. The flow diagram of included and excluded records at each step is presented in Figure 2.

Once the final corpus had been compiled, the metadata were used to perform bibliometric analysis using two complementary tools:

Biblioshiny, developed as an interactive interface to the R package (Version 4.4.2) Bibliometrix, stands out for its rich functionality. It offers a complete suite of statistical techniques combined with advanced visualization capabilities, enabling both analysis of scientific performance (number of publications, citations, productivity by author or source) and conceptual mapping of the field of study [75,76]. As an open-source tool, it is widely used in recent bibliometric studies for its flexibility and accessibility.

In a complementary vein, VOSviewer (version 1.6.20), developed in 2010 by Nees Jan van Eck and Ludo Waltman at Leiden University [77], is a software package specialized in the construction and exploration of bibliometric maps based on co-occurrence, co-citation, or collaboration data. It is renowned for the visual quality of its graphs and its ability to represent relationships between authors, keywords, or sources in the form of networks [78]. Its popularity in recent studies is due in particular to its compatibility with different databases and its ability to process large volumes of data efficiently [79].

3. Results and Discussion

3.1. Bibliometric Analysis

The bibliometric analysis conducted on the selected corpus reveals several key features of the scientific activity focused on energy optimization in electric vehicles. This section offers a structured interpretation of the extracted data, relying on standard bibliometric indicators and cartographic visualizations produced using VOSviewer (version 1.6.20),and Biblioshiny(for Bibliometrix, version 4.4.2). The results are presented along several complementary dimensions: the chronological evolution of publications, the distribution of sources and contributing authors, collaboration dynamics, and the dominant conceptual themes identified through keyword analysis.

3.1.1. Overview of the Analyzed Corpus: Annual Distribution of Publications (2017–2024)

From a temporal perspective, the annual distribution of publications selected for the bibliometric analysis, covering the period from 2017 to 2024 (Figure 3), reveals a significant evolution in the level of interest shown by the scientific community toward this topic. The number of publications per year is based on the curated dataset obtained from Scopus using the methodology described in Section 2 (Table 1 and Table 2). In 2017 and 2018, the number of articles remained low (6 and 3 publications, respectively), reflecting an early phase of emergence during which the subject began to draw attention but had not yet become the focus of sustained investigation [80,81].

Starting in 2019, a notable surge occurred with 16 publications [82], followed by consistently high levels in 2020 (15 articles) [83] and 2021 (14 articles) [84]. This upward trend highlights growing interest in issues related to energy optimization and electric vehicle range, likely driven by technological advancements in energy storage, power electronics, and control strategies. The year 2022 marked a peak with 23 publications [85], indicating a heightened enthusiasm for the subject. This increase can be attributed to a combination of factors, including intensified pressure for energy transition, accelerated decarbonization policies in the transport sector, and the delayed publication of research initiated during the pandemic period.

In contrast, a slight decline is observed in 2023 (12 articles) [86], which may signal a phase of stabilization or a shift in focus toward more specific or experimental issues. Finally, 2024 shows 13 publications [87]. This figure can reasonably be seen as indicative of a potential recovery, though the full scope of this trend will only become clear with the benefit of hindsight in the coming years.

Overall, this temporal evolution reinforces the relevance of the selected corpus, underscoring the gradual emergence of a now-structured and scientifically active research field.

3.1.2. Scientific Mapping of Journals

To identify the main publication sources within the field of energy optimization for electric vehicles, a journal analysis was conducted based on the frequency of publications among the selected articles. The results show that publications are distributed across a wide range of journals, reflecting the growing interest in this research domain. This editorial mapping helps to highlight the most influential journals in the field and to guide future research efforts toward the most relevant sources. Figure 4 shows the distribution of publications per journal across the entire dataset, while Figure 5 illustrates the temporal evolution of journal-specific scientific output from 2017 to 2024. The journal Energy stands out with 11 publications, confirming its status as a leading reference in the field of energy transition and energy efficiency as applied to transport systems. This multidisciplinary journal covers topics ranging from energy modeling to sustainable policy management, making it a highly suitable platform for research on consumption optimization strategies in electric vehicles. IEEE Access, with 8 publications, also plays a significant role. It is known for its broad coverage of technological domains, particularly those related to artificial intelligence, embedded systems, and automation, key dimensions for advanced control strategies in electric vehicles. IEEE Transactions on Vehicular Technology and World Electric Vehicle Journal follow, each with 5 publications. The former specializes in technologies for intelligent vehicles, connectivity, and energy management in transportation systems, while the latter focuses specifically on electric, hybrid, and hydrogen vehicles, offering a dedicated platform for innovation in e-mobility. Journals such as Energies, Journal of Energy Storage, and Applied Energy complement this core group, with 3 to 4 publications each. Energies is an open-access journal highly active in areas related to modeling, simulation, and optimization of energy systems. Applied Energy is recognized for the rigor of its contributions and its integrated systems perspective, while the Journal of Energy Storage targets battery technologies, a key subsystem for electric vehicle range. The presence, though more moderate, of journals such as Energy Conversion and Management, IEEE Transactions on Industrial Informatics, Control Engineering Practice, and Engineering Applications of Artificial Intelligence underscores the cross-disciplinary nature of this research topic.

3.1.3. Analysis of the Most Influential Authors

The bibliometric examination of authorship provides key insights into the most active and influential contributors in the field of energy optimization for electric vehicles. By combining quantitative and temporal indicators, the analysis highlights not only patterns in scientific productivity but also the underlying collaboration networks that structure this research domain.

Figure 6 displays the most prolific authors based on their number of publications within the top 100 articles selected for analysis. At the forefront, Wang L. stands out with nine publications, followed by Wang Y. and Wang J., with seven and six papers, respectively. Other frequent contributors include Li J., Li Y., Zhang Q., Li H., and Liang J., each with four to five publications. The prevalence of Chinese names underlines the prominent role played by researchers from China in advancing this field.

This productivity also reveals a temporal dynamic, as shown in Figure 7, which illustrates the evolution of individual authors’ publication activity over the years. Certain researchers, such as Li J., have maintained a sustained contribution from 2017 to 2024, reflecting long-term engagement in the topic. In contrast, the emergence of newer contributors in recent years signals a progressive renewal of the research landscape, potentially bringing novel approaches and perspectives.

Beyond individual productivity, Figure 8 presents a co-authorship network generated with VOSviewer, where each node represents an author and its size is proportional to the number of publications. Links between nodes indicate co-authored publications, forming distinct collaboration clusters differentiated by color. Notably, authors like Wang L. occupy central positions within the network, suggesting not only high productivity but also a key role in mobilizing collaborative research teams. This network-based visualization sheds light on the structural dynamics of knowledge production and reveals active hubs of expertise that are instrumental in driving innovation and dissemination across the field.

3.1.4. Author Affiliations

The analysis of author affiliations (

Table 3. Most Relevant author affiliations.

Rank	Affiliation	Articles
1	Jiangsu University	27
2	School of Mechanical Engineering	19
3	Chongqing University	18
4	Jilin University	15
5	Tsinghua University	13
6	Liaoning University of Technology	12
7	Beijing Institute of Technology	11
8	Central South University	11
9	University of Science and Technology Beijing	10
10	University of Science and Technology of China	9

) reveals a marked concentration of publications within Chinese institutions. Jiangsu University stands out clearly at the top with 27 articles, underscoring its strong, structured commitment to research in this field. The School of Mechanical Engineering follows with 19 publications, highlighting the central role of mechanical engineering departments in addressing energy management challenges for electric vehicles.

Chongqing University (18 articles), Jilin University (15 articles), and Tsinghua University (13 articles) also demonstrate sustained scientific output. These universities are internationally recognized for their work on transport electrification, embedded control systems, and advanced energy management strategies.

The remainder of the top 10 most productive institutions includes Liaoning University of Technology, Beijing Institute of Technology, Central South University, University of Science and Technology Beijing, and University of Science and Technology of China.

This distribution confirms the existence of true centers of excellence dedicated to electromobility research, reflecting a particularly strong and coordinated research dynamic in China around electric vehicle energy challenges. It also emphasizes the strategic importance of these academic hubs in advancing innovative solutions to improve energy efficiency, sustainability, and competitiveness in the global EV market.

3.1.5. Scientific Mapping of Contributing Countries

To determine the geographical regions most actively engaged in the field of energy management and optimization for electric vehicles, two complementary visualizations were employed. The first figure (Figure 9) displays the international collaboration network between countries: each node represents a country, with its size reflecting the volume of scientific output, while the thickness of the connecting lines indicates the frequency of bilateral collaborations. The second (Figure 10) provides a density map of scientific publications, highlighting regions with a high concentration of scholarly activity.

In both visualizations, China emerges as a central contributor, leading not only in terms of publication volume but also in the extent of its international collaborations. Alongside China, other countries such as the United States, India, and the United Kingdom also occupy significant positions within the global research network. Although certain collaborations appear less frequent, they nonetheless illustrate a gradual expansion of international scientific cooperation in this domain.

This geographical distribution is consistent with the author-level analysis, which identified several Chinese researchers among the most influential contributors. The predominance of China can be explained by a combination of structural and strategic factors. It reflects the size and rapid growth of the Chinese EV market, which is estimated at USD 357.98 billion in 2025 and is projected to reach USD 788.20 billion by 2030 [88]. Government incentives and national R&D programs, directly targeting the development of electric vehicles, battery technologies, and energy management solutions, further reinforce this leadership [89]. The share of Chinese publications in the field of EV batteries accounts for about 20% of global scientific output, according to a study by the Information Technology and Innovation Foundation (ITIF) [90]. This position is also strengthened by the presence of a complete industrial ecosystem, ranging from battery manufacturers (e.g., CATL, BYD) to automotive OEMs, fostering close collaboration between academia and industry. In addition, Chinese researchers benefit from access to large-scale experimental platforms: in 2023, the cumulative production of EV batteries in China exceeded 293.6 GWh, with a year-on-year growth of around 36.8% [91].

Taken together, these elements position China as a major hub for electric vehicle research [92] and explain the strong visibility of Chinese authors in the scientific literature on energy optimization for EVs.

3.1.6. Keyword Analysis

Keyword analysis constitutes a fundamental step in any bibliometric study, as it provides preliminary insights into the dominant research themes shaping a scientific field—without delving into the detailed content of individual contributions. Two visualizations were employed in this section: a bar chart displaying the ten most frequently occurring keywords, which highlights the core concepts of the dataset (Figure 11), and a co-occurrence network generated using VOSviewer that illustrates the semantic connections and thematic groupings among these keywords (Figure 12). Together, these visual representations offer a complementary perspective by combining frequency data with relational structures, allowing for the formulation of initial hypotheses about the intellectual organization of the field.

• Keyword Frequency: Core Concepts of the Dataset

The frequency analysis reveals several highly recurring terms, including energy management (52 occurrences), electric vehicles (39), energy management systems (34), secondary batteries (32), energy storage (28), energy utilization (27), control strategies (25), and electric machine control (24). This lexical concentration points to a clearly defined research orientation: optimizing energy usage in electric vehicles through intelligent management approaches, efficient use of on-board resources, and the development of advanced control strategies. These results also validate the relevance and consistency of the selected dataset.

• Keyword Co-Occurrence Map: Thematic Clusters and Emerging Structures

The co-occurrence map further enriches the analysis by revealing how these keywords interact within the research landscape. Each node in the network represents a keyword, with its size indicating frequency and its proximity to others signifying co-occurrence strength. Colors distinguish automatically identified semantic clusters, whose interpretation offers insights into underlying thematic groupings. This clustering highlights the emergence of structured subdomains within the broader research area, offering a more nuanced understanding of its internal dynamics.

Red Cluster: This cluster is characterized by the large size of its nodes, including keywords such as energy management, energy storage, secondary batteries, supercapacitors, hybrid energy storage systems, and HESS. The density and strong interconnectivity of this group suggest a substantial body of research focused on energy management within architectures that integrate multiple storage sources. The presence of terms such as fuzzy inference, hybrid energy storage system (HESS), and real-time control strategy indicates a growing interest in intelligent control approaches and real-time optimization. The significant size of the nodes suggests that these topics are already well-established within the literature.

Blue Cluster: Dominated by keywords such as battery management systems, temperature control, thermal management, air conditioning, energy savings, energy conservation, energy consumption, control strategies, and energy utilization, this cluster reflects a focus on battery energy management. It encompasses thermal aspects and the regulation of energy flows under varying operational conditions. The intermediate size of the nodes implies that while this thematic area is present in the literature, it remains subject to ongoing and diverse research efforts.

Green Cluster: This group features keywords such as electric machine control, traction motors, dual-motors, sliding mode control, vehicle wheels, electric drives, torque distribution, longitudinal control, permanent magnets, vehicle performance, dynamics, stability, and control systems. The cluster represents research oriented toward the optimization of motor control under different configurations (e.g., synchronous machines, dual-motor systems). The inclusion of techniques such as sliding mode control points leads to the use of advanced, potentially robust or predictive control methods. The high frequency of certain terms suggests that this domain constitutes a key technical pillar within the dataset.

Yellow Cluster: This cluster is led by terms such as regenerative braking, braking energy recovery, charging (batteries), electric discharges, energy recovery, adaptive control systems, dynamic programming, and deep learning. It reflects a thematic focus on harnessing deceleration phases for energy recovery, often in conjunction with numerical optimization techniques or artificial intelligence. The intermediate node sizes and the cluster’s relatively central position in the map indicate that this is a dynamically evolving area of research, potentially overlapping with other domains in the field.

Purple Cluster: More compact yet clearly distinct, this cluster includes terms such as driving pattern, driving pattern recognition, pattern recognition, fuzzy logic control, particle swarm optimization, electric loads, adaptive control systems, and vehicle-to-vehicle (V2V). The keywords suggest a research orientation toward driving behavior analysis and the integration of environmental or inter-vehicle data. Although the nodes are smaller in size, their high connectivity with other clusters indicates that these approaches are increasingly being incorporated into broader energy management strategies.

Each cluster reveals distinct technical and conceptual logics, with their richness reflected in both the diversity of terms and their spatial relationships. These results provide a robust foundation for the thematic analysis phase, in which the detailed examination of article content will help confirm, refine, or reorganize the hypotheses derived from this semantic mapping.

3.2. Thematic Approach to the Literature

The study presented here is based on a systematic analysis of a selected corpus of scientific publications to clarify how the literature addresses the issue of energy consumption optimization in electric vehicles (EVs). The findings reveal that current research does not focus on a single technological aspect but instead reflects a diversity of approaches, all converging toward the same overarching goal: increasing vehicle range by minimizing energy losses, while meeting key requirements in terms of safety, comfort, and durability.

A thorough review of the literature has led to the identification of several priority areas of intervention, each linked to one or more critical subsystems within the EV. These areas are not treated in isolation; rather, they operate within a cohesive framework, where energy efficiency emerges from the dynamic interplay between internal energy flows (electrical, thermal, mechanical) and the external conditions under which the vehicle operates. To facilitate a better understanding of this systemic complexity, Figure 13 provides a synthetic representation of the functional and energetic interactions between the main onboard subsystems and their operating environment. It highlights the different types of energy flows, electrical, mechanical, and thermal, and the cross-interactions between internal components and external constraints. Each subsystem can, depending on its role, consume, convert, or recover energy while remaining in continuous interaction with other parts of the vehicle.

At the core of the system lies the traction subsystem, which encompasses the electric motor, power electronics, and drivetrain. This assembly is responsible for converting electrical energy into mechanical energy to ensure propulsion [93]. The energy required for this process is supplied by the storage subsystem, primarily made up of batteries, and in some cases complemented by supercapacitors, which help support transient power demands and supply auxiliary systems [94]. Efficient management of this storage is essential for smoothing power peaks, extending battery lifespan, and maintaining stable energy delivery [95]. In parallel, the thermal management subsystem handles both cabin climate control and the regulation of critical component temperatures (battery, motor, power electronics, etc.) [96]. Its energy consumption is highly sensitive to environmental conditions, which makes intelligent control strategies crucial for preserving both vehicle performance and driving range. These three subsystems do not operate in isolation. The figure also emphasizes their connection to two major external environments. On one hand, the electrical grid acts as the primary energy source during charging and, in the context of Vehicle-to-Grid (V2G) integration [97], may also receive energy fed back by the vehicle. On the other hand, the road and climatic environment directly impact thermal loads, power demand, and vehicle dynamics [98].

This comprehensive view helps explain why researchers are moving away from isolated component-level optimization. Instead, they are embracing a systems-based approach in which improved range results from the coordinated management of heterogeneous energy flows, orchestrated across multiple functional layers.

The contributions identified in the literature can be grouped into six main thematic axes (Figure 14). The first two focus on the powertrain, with research dedicated to optimizing motor control, whether in single- or multi-motor architectures and enhancing energy recovery through regenerative braking.

The third axis addresses the intelligent management of hybrid energy storage systems, aiming to reduce battery stress and maintain stable power delivery. The fourth concentrates on integrated thermal management strategies, which become critical in demanding driving scenarios such as heavy climate control use, steep gradients, or fast charging.

Beyond these on-board dimensions, contemporary research is also expanding its scope to include external interactions. A fifth axis has emerged around inter-vehicle coordination, especially through cooperative driving strategies (V2V), where energy optimization is considered at the scale of a platoon or a convoy of vehicles. Finally, a sixth area of inquiry explores interaction with the electrical grid (V2G), examining how vehicles can not only draw energy but also actively contribute to the overall stability and efficiency of the broader energy system.

These six thematic areas (Figure 14) form the analytical foundation of this study and structure the developments that follow. They should not be viewed as isolated or self-contained domains, but rather as complementary levers that, each in its own way, contribute to reducing energy consumption and improving the driving range of electric vehicles.

A quantitative examination of our corpus highlights clear imbalances in the distribution of research efforts across these themes. Hybrid energy storage systems (HESS) and motor control, including torque distribution strategies, dominate the literature. This prominence is rooted in a combination of technical and practical factors. HESS directly addresses the energy–power trade-off by coupling a high–energy density battery, poorly suited to transient loads, with a supercapacitor or another high–power density device. Such architectures improve regenerative energy recovery and protect the battery from high current peaks, thereby extending its lifetime and enhancing effective vehicle range. Their benefits are straightforward to evaluate through both simulation and experimental testing, which facilitates validation and dissemination. In addition, HESS provides remarkable algorithmic flexibility: they accommodate heuristic rule-based control, optimization-based methods, and machine learning approaches, making them an attractive platform for diverse research contributions.

The strong focus on motor control and torque allocation reflects the central role of the powertrain in overall energy efficiency. Torque optimization directly influences electrical and mechanical losses, yielding immediate improvements in energy consumption. The maturity of control theory—spanning model predictive control, vector control, and adaptive regulation—together with robust modeling tools, has accelerated the development of advanced strategies, including those tailored for multi-motor architectures. In these cases, intelligent torque distribution across multiple machines represents a natural extension of classical control methods, opening up promising perspectives for improving not only efficiency but also dynamics and safety. A further advantage lies in the feasibility of implementation: many of these strategies can be deployed at low cost through software updates that leverage the vehicle’s existing power electronics, a factor that enhances their industrial appeal.

By contrast, other themes, while equally critical, remain less represented in the literature. Thermal management exemplifies this trend. It is fundamental to keeping components within their optimal operating windows and safeguarding vehicle autonomy, yet progress is hindered by the intrinsic complexity of multiphysics modeling—combining electrical, thermal, fluid, and chemical domains—and the high variability of real-world operating conditions. Regenerative braking faces a different issue: although it is a key mechanism for energy recovery, it is frequently embedded within broader EMS or HESS frameworks, and thus seldom appears as a standalone research focus.

Finally, emerging topics such as cooperative driving (V2V) and vehicle-to-grid interaction (V2G) reflect a paradigm shift toward systemic perspectives, in which the vehicle is no longer seen as an isolated unit but as an active node within a larger energy and mobility ecosystem. Nevertheless, these avenues remain embryonic. External dependencies, including communication reliability, standardization, infrastructure readiness, and regulatory frameworks, constrain their advancement. As a result, consolidated studies and large-scale experimental validations are still limited.

The following presents, in a structured way, a representative selection of articles from the corpus for each identified axis. For each theme, the strategies and methods proposed in the literature are described and clarified, providing a coherent overview of the approaches explored and their contributions to the energy optimization of electric vehicles.

3.2.1. Motor Control and Torque Allocation Strategies for Energy Efficiency Optimization in Single and Multi-Motor Electric Vehicle Architectures

The literature places significant emphasis on motor torque optimization, seen as a direct lever for enhancing the overall energy efficiency of electric vehicles. This issue lies at the heart of the powertrain’s functioning, where precise torque regulation enables mechanical energy to be aligned with the vehicle’s actual needs, internal losses to be minimized, and component lifespan to be extended.

Some approaches based on predictive models aim to anticipate future driving demands to generate an optimal torque in real time. For example, ref. [99] describes a Model Predictive Control (MPC) strategy that dynamically adjusts motor torque depending on the driver’s intent, whether sporty or economical, deduced from analysis of accelerator pedal behavior. The main goal is to develop an effective method to reduce energy consumption in electric vehicles while meeting drivers’ expectations. Similarly, ref. [100] applies a Finite Control Set Model Predictive Control (FCS-MPC) scheme that directly regulates motor currents to achieve precise control of torque and flux. FCS-MPC selects the optimal control action that minimizes a cost function while respecting the motor’s operational constraints. This approach offers real-time prediction and optimization of motor performance, reduces torque ripple, and adapts quickly to changing driving conditions, leading to more efficient energy use. Other research focuses more specifically on improving the electromechanical conversion efficiency itself. For instance, ref. [101] proposes an optimization strategy for the current angle in the electric drive system (EDS) using a golden section search method. This technique identifies the optimal angle across the full operating range and generates a lookup table of optimal values. Simulations that account for iron, copper, and inverter losses demonstrate that this strategy improves overall system efficiency under varying conditions, both in steady-state and transient regimes, compared to conventional control methods. Another notable trend is the integration of artificial intelligence to enhance control adaptability. Reference [102] introduces an intelligent PI controller with a neural observer that automatically tunes its coefficients based on operating conditions. This approach halves the transition time while reducing energy losses, inrush currents, and torque spikes. Meanwhile, ref. [103] presents a hybrid control strategy for the traction system of an electric vehicle with a permanent magnet synchronous motor (PMSM), combining neuro-fuzzy sliding mode control with an extended state observer (ESO). This approach significantly reduces chattering, improves disturbance rejection, and optimizes energy management, with simulations showing better tracking accuracy and overall performance compared to traditional methods. Some studies also target alternative motor architectures, such as switched reluctance motors. For example, ref. [102] highlights an Adaptive Supervisory Self-Learning Control (ASSC) strategy that combines neural networks, fuzzy inference systems, and first-order Takagi-Sugeno models to simultaneously regulate speed and torque. This method reduces ripples in speed, torque, current, and flux, adapts to varying loads, and achieves up to 95% efficiency compared to 85–91% for other tested methods. The result is lower losses, improved stability, and reduced energy consumption in electric vehicles.

While torque control in single-motor architectures already provides energy efficiency gains, the challenges become significantly more complex in the case of multi-motor electric vehicles. In such configurations, whether dual-motor systems, axle-distributed motors, or in-wheel motors, the issue extends beyond the individual control of each machine. The key lies in intelligently coordinating multiple motor sources to minimize overall energy consumption while respecting vehicle dynamics, safety, and ride comfort constraints.

Energy efficiency then depends on the system’s ability to optimally distribute torque among the motors, considering traction requirements, driving phases (acceleration, cruising, braking), grip conditions, and even road topography. This issue varies considerably depending on whether the vehicle uses a single-motor or multi-motor architecture.

Table 4. Comparative analysis, based on the analyzed literature, of single- and multi-motor electric vehicle architectures in terms of energy-related objectives, structural advantages, and implementation constraints.

Architecture	Main Objectives	Key Advantages	Limitations	References
Single-Motor	Maximize the main motor’s efficiency- Reduce electrical energy losses- Stabilize torque and dynamic response.	Simpler system structure → fewer components, lower parasitic losses Easier to model and control due to the centralized drive	– Limited flexibility in dynamic environments – No capability for distributed torque sharing – Mechanical constraints due to centralized powertrain	[99,100,101,102,103,104]
Multi-Motor	Optimize torque distribution across multiple motors Dynamically split torque between front and rear motors Switch between mono- and dual-motor modes (Dual-Motor) Ensure dynamic stability (yaw moment, traction) Minimize global drivetrain energy losses	Allows you to make the most of the high-efficiency zones of each motor. Fine-tune the power delivered by each engine according to its instantaneous efficiency. Better management of grip, gradients, and bends thanks to differentiated engine control. Flexible mode transitions Enhanced stability and comfort thanks to precise torque modulation on each axle or wheel.	– Complex coordination between motors: fine synchronization is necessary to avoid control conflicts or loss of efficiency. – Risk of imbalance in the event of a motor failure: the absence of active redundancy can compromise the stability or overall efficiency of the system. – High algorithmic complexity: multi-level control laws, particularly those incorporating prediction or optimization, increase the computational load. – High computational cost: online approaches require significant computational resources, which are sometimes not compatible with low-power embedded systems. – Risk of conflicting or redundant actions: In the absence of a robust supervision architecture, simultaneous decisions that are not harmonized between engines can occur.	[80,81,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119]

presents a comparative synthesis of these two configurations, highlighting the targeted energy objectives, structural advantages, and functional limitations, as analyzed in the recent literature.

Among the strategies employed in the analyzed works to address these challenges are:

• Hierarchical control schemes: These architectures are typically structured across several layers, with the upper layer defining global objectives (trajectory tracking, autonomy, comfort), while the lower layers handle torque allocation based on motor characteristics and dynamic constraints [80,107,110,113,115,116,117,118,119].
• Predictive control approaches: These methods anticipate short-term driving events (turns, braking, acceleration) to dynamically adjust torque distribution [105]. They may be coupled with databases or predefined operation maps.
• Optimization-based strategies: Techniques such as dynamic programming [81], bio-inspired metaheuristics (e.g., APSO) [108], and online methods such as the Equivalent Consumption Minimization Strategy (ECMS) [106] to improve the economy of the vehicle. They are used to determine the most efficient ways to use energy in electric vehicles along predefined routes, to reduce energy consumption.
• Robust and adaptive controls: In the presence of uncertainties (grip changes, mechanical disturbances, varying loads), approaches like adaptive sliding mode control (DASMC) [112] ensure stable torque regulation even under highly disturbed conditions. In a similar vein, article [111] introduces a robust torque control strategy for dual-motor electric drivetrains using planetary gear transmission, aiming to suppress jerks during mode transitions (e.g., switching between acceleration and regenerative braking). The control system combines PI-based speed feedback with a feedforward torque allocation logic, enhancing ride comfort and system stability despite drivetrain nonlinearities and switching disturbances.
• Intelligent and hybrid control strategies: The integration of artificial intelligence enables systems capable of learning optimal behaviors. Methods such as deep reinforcement learning (AD-DDPG) [109], neural networks trained on driving data [105], or fuzzy systems combined with evolutionary algorithms (like PSO) [114] allow dynamic torque distribution to be tailored to complex and varying driving conditions.

3.2.2. Regenerative Braking Strategies in the Context of Energy Efficiency Optimization for Electric Vehicles

Among the contributions identified in the analyzed corpus, several studies address regenerative braking as a core element in the broader context of energy efficiency improvement in electric vehicles (EVs). Far from being a mere energy recovery mechanism, regenerative braking plays a pivotal role in extending battery life and enhancing vehicle range. Unlike conventional braking systems, which convert kinetic energy into heat and dissipate it, EVs, when appropriately controlled, can redirect this energy back into the storage system. However, the full potential of regenerative braking can only be realized when it is integrated into a holistic energy management strategy that considers vehicle dynamics, battery state of charge (SOC), traction conditions, and safety constraints.

A significant number of studies highlight this integration as a key research avenue. Two primary approaches emerge in the literature. The first focuses on maximizing energy recovery during deceleration and braking phases while preserving the dynamic stability of the vehicle. The second seeks to reduce the impact of regeneration on battery health by avoiding current spikes and intelligently coordinating the load between mechanical and electrical braking. In this context, article [120] proposes an intersection-aware energy management strategy designed specifically for connected electric vehicles. Recognizing the considerable energy losses associated with stop-and-go behavior at traffic signals and intersections, the study emphasizes the importance of optimizing braking phases in such urban scenarios. By integrating regenerative braking into the energy management strategy, the vehicle can recover a substantial portion of energy otherwise lost through conventional braking. The authors introduce two control approaches: a rule-based strategy and a fuzzy logic controller, both aimed at enhancing energy recovery while ensuring smooth and efficient vehicle deceleration at intersections.

Several studies in the dataset adopt advanced strategies, including adaptive, predictive, or intelligent control schemes capable of adjusting the recovered power in real time according to driving conditions. For example, article [121] introduces a fuzzy logic–based strategy that dynamically modulates the share of regenerative braking depending on the battery’s state of charge and deceleration demand. This approach aims to optimize energy recovery while avoiding overcharging by adapting the braking command to the vehicle’s real-time context. Similarly, article [122] combines fuzzy control with adaptive sliding mode control to enhance braking performance. This hybrid approach targets braking torque regulation by accounting for road conditions and vehicle behavior, notably through optimization of wheel slip ratios. The reported experimental results demonstrate substantial improvement, with energy recovery increasing from 83.23 to 88.09 kJ compared to a standard sliding mode strategy. Another study [123] focuses on the regenerative braking control of pure electric mining dump trucks, a domain with particularly harsh and highly variable operating conditions due to large fluctuations in load and road slope. The study introduces an adaptive energy management framework leveraging deep reinforcement learning, specifically the Soft Actor–Critic (SAC) and Deep Deterministic Policy Gradient (DDPG) algorithms. The model considers vehicle speed, acceleration, mass, road gradient, battery state of charge (SOC), and charge–discharge rate as state variables, while gear shifting serves as the control variable. To enhance adaptability under varying operating conditions, an automatic entropy tuning mechanism is incorporated. Simulation outcomes indicate that, in comparison with a rule-based control scheme, the proposed SAC- and DDPG-based optimization strategies, as well as the dynamic programming approach, achieve respective gains in energy efficiency of 18.15%, 17.18%, and 16.63%. In addition, battery lifespan increases by 57.31%, 56.87%, and 57.38%, respectively. These results highlight the potential of deep reinforcement learning to optimize regenerative braking in heavy-duty EV applications with complex, dynamic energy demands. Other contributions apply predictive strategies that anticipate future driving conditions to enhance energy planning. Article [124], for instance, integrates a supervised learning algorithm (C4.5) with a long short-term memory (LSTM) neural network to forecast driving behavior. The predicted information is then used by the Seagull Optimization Algorithm to manage torque distribution and improve targeted regenerative recovery. Another relevant direction involves personalizing regeneration based on driver-specific patterns. As demonstrated in article [125], a strategy using iterative dynamic programming (IDP) coupled with a bidirectional LSTM network (BLSTM) adjusts energy recovery to the driver’s preferences, enhancing both braking smoothness and energy efficiency by aligning system behavior with human input. Studies even highlight the importance of coordination between hydraulic and regenerative braking. Article [126] develops a coordinated braking control strategy for electric vehicles, aiming to balance energy recovery and battery lifespan. The approach is based on detailed modeling of braking energy consumption and battery degradation. A multi-objective model predictive controller (MPC) is designed to optimally distribute braking force between motor and hydraulic systems. The method is validated through AMESim/Simulink co-simulations, showing how different braking intensities and initial SOC levels affect both energy recovery and battery aging.

The summary table (

Table 5. Summary of Identified Approaches in the Analyzed Literature for Energy-Efficient Regenerative Braking Control.

Ref.	Specific Method	Objective
[120]	Fuzzy Inference System (FIS) integrated into a rule-based energy management strategy.	Optimize energy recovery at urban intersections.
[121]	Real-time adaptive fuzzy control.	Dynamically adjust regeneration level based on battery SOC and deceleration.
[122]	AFSMC (Adaptive Fuzzy Sliding Mode Control).	Maximize energy recovery while regulating wheel slip.
[123]	Deep Reinforcement Learning: SAC and DDPG with entropy tuning.	Optimize regenerative braking under varying load and slope
[124]	Hybrid AI and Bio-Inspired Optimization (C4.5 LSTM Seagull Optimization Algorithm).	Adjust regeneration based on predicted route and driving behavior.
[125]	Deep Learning and Optimization: IDP-BLSTM (Dynamic Programming + BiLSTM).	Adapt regeneration to individual driving style.
[126]	A multi-objective model predictive controller (MPC).	Manage regeneration based on battery SOC and safety constraints.

) above presents a synthesis of the various approaches discussed in the reviewed literature, highlighting the diversity of strategies aimed at optimizing regenerative braking in EVs for improved energy performance.

3.2.3. Intelligent Energy Management in Hybrid Energy Storage Systems (HESS) for Enhanced Efficiency and Battery Longevity

A research focus that stands out in the analyzed works concerns the development of energy management strategies aimed at dynamically and contextually orchestrating the distribution of power between different energy sources. This issue is especially relevant in hybrid energy storage systems (HESS), which typically combine a battery serving as the main source with high energy density but poorly suited to transient demands with a supercapacitor or another high-power-density element capable of absorbing or supplying instantaneous peaks.

This is a particularly well-established topic in the literature, as evidenced by the large number of contributions identified in our corpus. It was clearly revealed from the bibliometric analysis phase, through the emergence of keywords at the core of the densest cluster in the co-occurrence map. The main objective is to optimize energy exchanges between these two complementary sources in order to extend battery lifespan, reduce energy losses, and ensure greater stability of the propulsion system, especially on the DC bus. However, the actual effectiveness of the HESS relies on the adopted energy management strategy.

Before addressing energy management strategies, many studies first focused on low-level dynamic control of HESS systems. These strategies, derived from advanced nonlinear control, aim to guarantee stability, precision, and robustness of the system in the face of disturbances and load variations. Among the approaches used are adaptive Sliding Mode Control (SMC) [127], Lyapunov-based control [128], H∞ laws [129], techniques based on differential flatness [130], as well as alternative converter and inverter topologies specifically designed to improve current sharing in HESS configurations [131].

Through the analysis of the reference articles, we identified the most commonly used approaches and analyzed their direct contribution to energy efficiency goals. These approaches are classified in the following categories:

Rule-based methods (deterministic or fuzzy):

These approaches are the most widespread in the early development phases. They rely on predefined rules, often derived from human expertise or simplified models. Heuristic rules [83,132,133,134,135] and fuzzy logic [136,137,138,139,140] enable management of energy transfer by activating the supercapacitor during transient phases (acceleration, braking), thus preventing the battery from being subjected to excessive current demands. These strategies ensure operation within the battery’s optimal efficiency range, thereby reducing overall energy consumption over a complete cycle.

Optimization approaches (offline or real-time):

Optimization strategies, such as Model Predictive Control (MPC) [141,142,143,144,145], Dynamic Programming (DP) [146], or meta-heuristic algorithms (such as PMP, PSO, GA) [84,147,148,149], aim to anticipate energy demand over a given time horizon and plan the optimal distribution between battery and supercapacitor. By minimizing overall energy consumption and reducing power peaks absorbed by the battery, these methods help extend the vehicle’s effective range. Among the optimization strategies, some deterministic approaches based on analytical solutions to multi-objective problems have also been proposed. For example, a strategy formulated as a multi-objective control (MOO) problem—combining loss reduction, system stability, and battery preservation—is solved using Karush-Kuhn-Tucker (KKT) conditions, enabling fast real-time implementation without requiring demand anticipation [150]. This approach stands out for its ability to provide an optimal solution with reduced complexity, unlike dynamic programming or predictive methods.

Machine learning and hybrid strategies:

More recently, research has turned toward integrating artificial intelligence techniques (neural networks, reinforcement learning, etc.) into energy management strategies to endow control systems with learning, adaptation, and prediction capabilities [82,151,152,153]. These techniques leverage data from the vehicle and its environment to anticipate energy needs, adjust to varying driving conditions, and optimize power distribution in real time.

At the same time, there is a strong trend toward hybrid strategies [87,154,155,156,157,158,159,160,161,162,163,164] that combine the complementary advantages of the previously developed approaches: heuristic rules, deterministic or stochastic optimization, predictive control, fuzzy logic, or AI.

The various studies analyzed are summarized in the following

Table 6. Summary of Energy Management Strategies for HESS in Electric Vehicles: Methods, Benefits, and Implementation Constraints Identified in the Analyzed Literature.

Ref	Title	Method	Results	Constraints
[127]	“Implementation of an estimator-based adaptive sliding mode control strategy for a boost converter-based battery/supercapacitor hybrid energy storage system in electric vehicles,”	Adaptive SMC enriched by an estimator	Highly robust in the face of uncertainties and disturbances. Battery preservation (thanks to fine-tuned current management). System stability even under variable conditions.	Complex design (integration of adaptability and estimation). Higher computation cost in real time. Dependence on estimator quality (critical accuracy).
[128]	“Sliding-mode and Lyapunov function-based control for battery/supercapacitor hybrid energy storage system used in electric vehicles,”	Combined Lyapunov + Sliding Mode Control (SMC)	Stability guaranteed by the Lyapunov function. Dynamic energy sharing between the battery and SC. Efficient maintenance of DC bus voltage.	Sensitive to inaccurate models (especially for Lyapunov). Simple EMS: may not be optimal in all driving situations.
[129]	“H∞ Control System for Battery and Supercapacitor Hybrid Energy Storage in Electric Vehicles”	Robust H-infinity controller	Precise control of traction and regeneration current. Reduced complexity thanks to the second-order model. Optimized power sharing in both phases (traction/braking).	Simplified model: possible loss of accuracy in certain dynamic speeds. Controller is sensitive to parameter settings. No explicit adaptability to real driving conditions.
[130]	“Differential Flatness-Based Cascade Energy/Current Control of Battery/Supercapacitor Hybrid Source for Modern e–Vehicle Applications,”	Flatness Control (Model-based flatness control)	This strategy enables several variables (currents, DC bus energy, supercapacitor, etc.) to be controlled simultaneously. Better dynamics than conventional PI control Simplicity of control once the “flat outputs” have been identified-reduced complexity of control laws for complex systems.	Requires trajectory planning: mandatory to implement the control law. Accurate modeling is required to identify flat outputs. Not always applicable: not all systems are differentially flat.
[132]	“Global sliding mode control of vehicle-mounted hybrid energy storage system based on exponential reaching law”	Rule-based with a global sliding mode control strategy based on exponential reaching law (E-GSM)	Optimized distribution of energy between battery and supercapacitor. Improved robustness against disturbances, better reference tracking dynamics. (E-GSM) Guaranteed bus voltage stability.	Rule-based EMS: non-optimal and non-adaptive. E-GSM is sensitive to parameter choices and extreme conditions.
[133]	“Pioneering Battery-Supercapacitor Hybrid Energy Management for E-Scooters for Sustainable Urban Transportation,”	bang-bang control	Reduced stress The battery extends its life improve overall system efficiency	Simple but not adaptive
[83]	“Optimal Control Strategy to Maximize the Performance of Hybrid Energy Storage System for Electric Vehicle Considering Topography Information,”	Adaptive fuzzy control with CPS (contour positioning system) integrated into rule-based EMS	Intelligent power distribution according to the slope of the terrain, for greater energy efficiency and reduced stress on the battery.	Dependence on slope measurement accuracy; complexity of implementing adaptive controllers in real-life conditions.
[134]	“Composite Non-Linear Control of Hybrid Energy-Storage System in Electric Vehicle,”	a rule-based energy management strategy	Good robustness against system uncertainties. Precise tracking of current references. Optimum power distribution between sources (battery/SC). Stable DC bus voltage. Allows efficient optimization of EV power under variable conditions.	Complexity of implementation (combination of several nonlinear methods). Depends on model quality for exact linearization.
[136]	“Adaptive power allocation strategy for hybrid energy storage system based on Hilbert-Huang transform,”	Fuzzy control with frequency analysis	Dynamic frequency adaptation of energy flow Reduced battery stress	Complex tuning, possible latency
[137]	“Vehicle Speed Optimized Fuzzy Energy Management for Hybrid Energy Storage System in Electric Vehicles,”	Speed-based fuzzy control (VSO-FEMS: vehicle speed optimized fuzzy energy management strategy)	Energy saving and longer battery life reduces total energy consumption by 1.28% compared to a conventional fuzzy control strategy	Strong dependency on speed prediction
[138]	“Research on vehicle controller and control strategy for pure electric vehicle with composite power,”	fuzzy logic	intelligent, flexible management of energy flows according to driving scenario Reduces the complexity of control based solely on rigid equations. Easy to adapt to new cases with only added rules.	Requires good tuning
[139]	“Multi-Fuzzy Control Based Energy Management Strategy of Battery/Super-capacitor Hybrid Energy System of Electric Vehicles.”	Fuzzy control	Peak reduction Battery protection Improved stability and efficiency	Complexity implementation multi-controller
[140]	“Research on Energy Management Strategy of Battery-Flywheel Hybrid Energy Storage Electric Vehicle,”	Fuzzy control	Lower battery stress, inertia-based energy support	complex mechanical integration
[141]	“Implementation of a predictive energy management strategy for battery and supercapacitor hybrid energy storage systems of pure electric vehicles,”	Markov chain-based power demand prediction with fuzzy logic control.	Current peak reduction Stable current flow	Rigid filtering, limited flexibility
[142]	“A Power Distribution Strategy for Hybrid Energy Storage System Using Adaptive Model Predictive Control,”	AMPC Adaptive MPC	More effective in terms of system efficiency and battery conservation, as the peak battery cell current can be reduced by at least 24.4% and the total energy loss by at least 6.4% with the proportional integral (PI) and model predictive control methods. Battery amp-hour output and root-mean-square battery current can be reduced by 16.2% and 29.8%, respectively.	Design complexity High computational load Sensitivity to the quality of the adaptive model: errors in adaptation can degrade performance.
[143]	“A Predictive Set-Point Modulation Energy Management Strategy for Hybrid Energy Storage Systems,”	Predictive Modulation (predictive setpoint modulation)	better use of the supercapacitors, reducing the load on the battery. Protects battery life. Reduced voltage fluctuations on the DC bus, ensuring stable operation. Long-term sustainable power supply thanks to adaptive power allocation.	Complexity of implementation Dependence on the system model and increased computational load Difficulty of optimization under variable conditions
[144]	“A predictive power management scheme for a hybrid energy storage system in an electric vehicle,”	Modified MPC (Simplified MPC for DC bus voltage)	Stable bus regulation Reduced computation	Simplified model
[145]	“Linear Parameter-Varying Model Predictive Control for Intelligent Energy Management in Battery/Supercapacitor Electric Vehicles,”	LPV-MPC (Linear Parameter Varying MPC)	Dynamic energy distribution Accurate and predictive	Model uncertainty sensitivity
[146]	“Load-adaptive real-time energy management strategy for battery/ultracapacitor hybrid energy storage system using dynamic programming optimization,”	dynamic programming optimization DP	reduces battery wear and tear (reduction in total current from 3.4% to 15.7%) and energy losses (reduction from 3% to 15.1%), compared with a conventional per-rule strategy	Non-adaptive rules are fixed after offline training
[147]	“Multi-objective benchmark for energy management of dual-source electric vehicles: An optimal control approach,”	alt-PMP (alternative Pontryagin’s Minimum Principle	minimizing battery degradation and SC subsystem losses.	Choice of critical weighting coefficients: incorrect settings can unbalance objectives. Less robust than AI methods in the face of unmodelled uncertainties or unforeseen driving variations.
[148]	“A significant energy management control strategy for a hybrid source EV,”	(Particle Swarm Optimization—PSO)	Efficiently finds the best way to share power between the battery and the supercapacitor, to improve the vehicle’s range and performance while preserving battery life.	Not adaptive in real-time
[149]	“Combined Sizing & Energy Management of Hess for An Electric Vehicle by PSO With Novel Power Sharing Control Strategy”	Particle Swarm Optimization (PSO)	Reduced consumption	Lack of dynamic flexibility Complex multi-goal setting
[84]	“Optimization and control of battery-flywheel compound energy storage system during an electric vehicle braking,”	Genetic algorithm	Intelligent distribution of energy between the two sources, depending on driving conditions 42.27% reduction in maximum battery charging current compared with a single battery system, reducing wear and improving service life. Improved stability and robustness of the flywheel system	Computationally intensive genetic algorithm The need for a precise model
[150]	“Multi-objective optimization-based real-time control strategy for battery/ultracapacitor hybrid energy management systems,”	Multi-objective optimization solved using the Karush-Kuhn-Tucker	Reduces energy loss and battery stress. Extends battery life by limiting overcharging. Improves overall hybrid system performance thanks to better coordination between the two sources.	Requires fine calibration of weights Dependence on precise models
[151]	“A soft actor-critic-based energy management strategy for electric vehicles with hybrid energy storage systems,”	Soft Actor Critic RL	The results of this research indicate that SAC-based EMS reduces HESS energy loss by 8.75% and 6.09% compared to Deep Q Network (DQN)- and Deep Deterministic Policy Gradient (DDPG)-based EMS, respectively. Stable and efficient thanks to offline learning Less sensitive to hyperparameters Performs complex continuous-action tasks	More computationally expensive as it uses multiple networks May converge slowly due to continuous exploration
[82]	“Energy management of an electric vehicle by a hybrid energy storage system with novel control strategy,”	neural networks (ANN)	SOC optimization and Voltage stability	Requires prior training
[152]	“Stochastic Control of Predictive Power Management for Battery/Supercapacitor Hybrid Energy Storage Systems of Electric Vehicles,”	neural networks (ANN)	Efficient anticipation of energy demand Reduced battery wear through frequency separation Adaptation to driving profiles	Requires training data for a neural network Sensitivity to prediction errors Complex implementation (combining prediction + optimization)
[153]	“Hierarchical Q-learning network for online simultaneous optimization of energy efficiency and battery life of the battery/ultracapacitor electric vehicle,”	Q-learning (Hierarchical tabular Q-learning)	Battery capacity conservation increased vehicle range Adaptation over time	Requires extensive training
[154]	“Intelligent Energy Management for Full-Active Hybrid Energy Storage Systems in Electric Vehicles Using Teaching-Learning-Based Optimization in Fuzzy Logic Algorithms,”	Teaching–Learning-Based Optimization with Fuzzy Logic	Energy reduction peak smoothing	Difficulty integrating into real-time environments
[157]	“Energy Management Control Strategy for Hybrid Energy Storage Systems in Electric Vehicles,”	Multilayer hybrid fuzzy (Fuzzy; Markov; Wavelet)	Adaptive energy allocation by dynamics Dynamic adaptation to demand	Difficult to tune, complex structure
[155]	“Optimization of Hybrid Energy Storage System Control Strategy for Pure Electric Vehicle Based on Typical Driving Cycle,”	fuzzy control strategy enhanced with GA	With GA, reducing energy consumption by 3 to 9% compared with the PSO, depending on the scenarios simulated.	Complex to tune
[158]	A real-time energy management control strategy for battery and supercapacitor hybrid energy storage systems of pure electric vehicles	Wavelet Transform (offline), Neural Network (trained to emulate wavelet decomposition), and Fuzzy Logic Controller	18% reduction in battery ageing costs compared with the conventional filtration-based control strategy Effective suppression of power peaks and reduction in battery current variations.	Need for offline pre-processing (neural network training from wavelet decomposition). Sensitivity to training quality. Combined hardware implementation complexity (neural network + fuzzy logic).
[159]	“Fuzzy Predictive Energy Management for Hybrid Energy Storage Systems of Pure Electric Vehicles using Markov Chain Model,”	Prediction-based fuzzy control (Fuzzy and Markov)	Smart energy allocation prediction Intelligent responsiveness	Sensitive to prediction uncertainty
[160]	“Optimal energy management for a Li-ion battery/supercapacitor hybrid energy storage system based on a particle swarm optimization incorporating Nelder-Mead simplex approach,”	Fuzzy and PSO/Nelder-Mead	Reduces stress on the battery, resulting in up to 20% longer battery life than a battery-only solution. Maintains good performance in terms of power output.	Algorithmic complexity (PSO and NM are non-trivial algorithms). Difficult to implement in real time without simplification or acceleration. Initial tuning required
[161]	“A regulatory power split strategy for energy management with battery and ultracapacitor,”	PI-regulated fuzzy control	Current peak suppression in HESS Good power distribution	Non-implementation specified on real cycles
[162]	“Spatial–temporal data-driven full driving cycle prediction for optimal energy management of battery/supercapacitor electric vehicles,”	LSTM-based velocity prediction and spatio-temporal interpolation (STIM) and multi-horizon MPC (MH-MPC)	Vision on several time scales: combines short and long horizons Anticipates changes in route, speed, or constraints, improving overall energy efficiency	Increased algorithmic complexity Need for reliable models over all horizons
[87]	“Optimal power-split of hybrid energy storage system using Pontryagin’s minimum principle and deep reinforcement learning approach for electric vehicle application,”	PMP with deep reinforcement learning (RL)	Minimize energy consumption and battery degradation Reduce aging and energy loss	High computation cost
[163]	“A new adaptive PSO-PID control strategy of hybrid energy storage system for electric vehicles,”	PSO-tuned PID	Significant improvement in performance compared to a conventional PID. Global optimization of energy management via PSO. Simple implementation with good stability and reduced energy consumption.	Offline approach: the PSO optimization phase is carried out offline, so there is no ability to adapt in real time. Less robust than advanced non-linear strategies (SMC, fuzzy, etc.) under highly variable or uncertain conditions.
[164]	“Fuzzy-Super Twisting control implementation of battery/super capacitor for electric vehicles,”	Fuzzy logic-based EMS and Super-Twisting Sliding Mode Control (ST-SMC)	Improved source longevity and system autonomy Smooth power distribution with dynamic adaptation to driving cycles Accurate DC-bus and SC voltage regulation Power frequency filtering: steady-state handled by battery, transients by SC Robust speed tracking with low torque/flux ripples	Rule-based EMS may need expert tuning for robustness in diverse real conditions Experimental validation limited to a small-scale prototype Integration of fuzzy logic with second-order SMC requires careful stability analysis
[156]	“Performance Analysis of MPBC with PI and Fuzzy Logic Controllers Applied to Solar Powered Electric Vehicle Application,”	Hybrid controller: MPBC (Measurement of Parameter-Based Controller) combined with fuzzy logic (FLC) and proportional-integral (PI) controllers, forming two different hybrid controllers named MPBC+FLC and MPBC+PI	Smooth transition between battery and supercapacitor based on motor speed and current. Improved power management under variable conditions.	Requires accurate real-time motor data (speed, current). Complexity due to the hybrid control structure. Dependence on the tuning of both MPBC and FLC/PI components.

, which highlights for each contribution the methods used, the results obtained, and the constraints considered.

This analysis shows that researchers do not simply aim to improve energy management in isolation. They are increasingly moving toward hybrid strategies that combine several methods capable of reconciling energy performance, robustness, and adaptability under real-world conditions, in order to leverage the specific advantages of each method while circumventing their limitations. The most recent trends integrate artificial intelligence, multi-objective optimization, and component aging criteria to pursue a common goal: making every joule of onboard energy more useful, thereby extending range without compromising safety or durability.

3.2.4. Thermal Management as a Core Lever for Enhancing Energy Efficiency in Electric Vehicles

Thermal management has emerged as a key area for improving the range of electric vehicles (EVs). Initially perceived as a concern limited to cabin comfort or component protection, it is now increasingly integrated into comprehensive energy optimization strategies. The studies reviewed in the analyzed database reflect this evolving perspective. They highlight the deployment of methods aimed at limiting thermal losses while ensuring dynamic performance, extended range, and vehicle safety. These investigations are generally based on a shared observation: under extreme climatic conditions, the demands for heating, air conditioning, or thermal regulation of batteries and other components can represent a substantial portion of the vehicle’s total energy consumption [165]. These energy demands directly reduce the effective driving range. Moreover, poor thermal management, whether due to excessive heat or cold, can lead to critical consequences. Under high temperatures, the risk of thermal runaway increases: the SEI (Solid Electrolyte Interphase) layer may degrade, triggering uncontrolled reactions that can potentially cause fires. Conversely, at low temperatures, lithium dendrite formation may occur, impairing electrochemical performance and, in some cases, irreversibly damaging the battery cells.

To address these challenges, the literature stresses the need for a systemic and integrated approach to thermal management. For instance, article [166] introduces a technical solution known as IVTM (Integrated Vehicle Thermal Management), which relies on multidimensional numerical simulations and collaborative multi-objective optimization. This approach aims to integrate subsystem coordination (batteries, electric motors, air conditioning, etc.) from the design phase to enhance thermal performance, dynamic control, and energy efficiency across all operating conditions. Other approaches focus on optimizing EV energy consumption by implementing alternative thermal management strategies. Article [167] introduces a strategy called TSEC (Two-Stage Eco-Cooling), which integrates predictive planning of target temperatures (for both cabin and battery) based on a multi-source analysis (weather conditions, vehicle speed, state of charge, passenger characteristics), followed by dynamic control using a fuzzy PID controller. The optimization of temperature trajectories is handled through Dynamic Programming (DP), while precise tracking significantly reduces thermal deviations. This strategy leads to a 21.48% reduction in battery capacity loss compared to an on-off controller, and an 8.55% reduction compared to a conventional PID controller, which translates into a significant slowdown in electrochemical aging. At the same time, the energy consumption of the thermal management system (TMS) decreases by 42.86% compared to on-off control, and 18.54% compared to PID, thereby freeing more energy for electric propulsion and directly contributing to increased vehicle range. Article [168] addresses another important aspect of direct-cooled battery thermal management systems under high-temperature conditions. It proposes adding a secondary throttle (ST) orifice at the refrigerant outlet of the battery branch to provide additional control flexibility, combined with a new control strategy based on a deep reinforcement learning (RL) algorithm. The RL controller simultaneously manages compressor speed and ST orifice opening to improve temperature regulation and reduce energy consumption. Results under uniform climbing and NEDC conditions show more stable control of the passenger compartment (22 °C) and battery cold plate (20 °C), while reducing compressor energy use by 5.7% and 7.3%, respectively, compared to a conventional PID strategy. This demonstrates the potential of advanced machine-learning-based control to enhance efficiency in challenging thermal environments.

Focusing more specifically on air conditioning, another contribution [169] implements a Model Predictive Control (MPC) scheme to control a transcritical CO₂ air conditioning system in real-time. Unlike TSEC, which aims to coordinate battery and cabin temperatures using fuzzy rules and global planning, MPC directly acts on physical actuators (compressor, expansion valve, fan) to maintain optimal conditions despite driving variations. Two controller variants are proposed: one prioritizing thermal comfort, the other aiming to minimize energy consumption. Experimental results over six hours of testing show a 20.27% energy saving (2.54 kWh) for the energy-oriented version, and 13.33% (1.67 kWh) for the comfort-oriented version, compared to a traditional PI control system.

Article [170] addresses another crucial aspect: the consideration of random load fluctuations and variable thermal demands due to air conditioning, which are often overlooked in conventional strategies. The authors propose combining driving pattern prediction (via a Markov chain) with fuzzy control capable of dynamically adapting energy management to anticipated disturbances. Results obtained under the NEDC cycle and validated through Hardware-in-the-Loop (HIL) simulations demonstrate an 8.14% improvement in battery energy performance compared to a logic-threshold-based strategy, reflecting better SOC maintenance and increased range, even under climatic and electrical perturbations.

Researchers are thus increasingly shifting towards approaches that aim to reduce internal losses, adapt thermal demands to contextual conditions, and enhance the overall efficiency of the vehicle.

3.2.5. Optimizing Energy Consumption in Cooperative Driving Strategies for Electric Vehicles

In the reviewed literature, a distinct research direction emerges around energy optimization in collective configurations of electric mobility (

Table 7. Summary Table—Energy Optimization Approaches in Cooperative Driving Strategies for Electric Vehicles Identified in the Analyzed Literature.

Article	Type of Cooperative Driving	Control Method	Optimized Parameters	Energy Optimization Strategy	Reported Energy Savings
[171]	Eco-Car-Following	Model Predictive Control (MPC)	Follower vehicle speed	Incorporation of power demand as a key term in the cost function	The proposed Economy-Oriented Car-Following Control (EOCFC) strategy achieved improvements in energy efficiency of 0.53%, 3.33%, and 1.51% under NEDC, UDDS, and WLTC driving cycles, respectively, compared to a standard multi-objective adaptive cruise control method
[85]	Eco-Car-Following	MPC combined with NARX prediction and variable weighting	Motor torque, vehicle speed, inter-vehicle distance	Minimization of motor energy consumption and mitigation of battery current peaks	Demonstrated superior efficiency and smoother torque control than classical MPC, LQR, PID, and dynamic programming under NEDC and WVUCITY conditions
[172]	Platooning with front- and rear-independent drive vehicles	Multi-objective Nonlinear MPC	Longitudinal control and torque distribution between axles	Optimization of total power consumption across the vehicle platoon	Reported overall platoon energy savings of 3.0%, 1.6%, and 4.7% under UDDS, WLTC, and HWFET test conditions, respectively
[173]	Platooning with constraint handling and energy-aware leader control	Distributed Reference Governor (RG) with Nonlinear MPC for the leader	String stability, leader energy consumption	Energy-efficient leader trajectory planning with RG constraint enforcement	Improved total energy economy of the platoon (exact gains not numerically specified), validated via HIL testing
[174]	Platooning with regenerative braking	Distributed MPC with brake force distribution strategy	Vehicle speed, trajectory, braking force	Optimized recovery of braking energy within regulatory safety limits	Improvement in energy efficiency validated through simulation and hardware-in-the-loop testing
[175]	Platooning with hybrid energy storage (HESS)	Distributed MPC (upper layer) and Rolling Horizon optimization (lower layer)	Speed profiles of each vehicle, battery, and supercapacitor energy allocation	Joint optimization of cooperative speed planning and real-time energy distribution in HESS systems	Exact numerical gain not specified, though effectiveness demonstrated through simulation results
[176]	Predictive platooning with terrain preview and V2V communication	Nonlinear MPC with terrain and leader motion prediction	Speed and motor torque	Anticipatory control using slope and leader trajectory forecasts to reduce energy demand.	Achieved superior efficiency compared to strategies without prediction, though quantitative gains were not explicitly provided
[86]	Cooperative eco-driving at intersections	Genetic Algorithm for offline speed planning and Sliding Mode Control for tracking	Predefined speed trajectory	Energy savings through smoothed speed transitions and V2X-based coordination	Up to 26% reduction in energy consumption compared to the rule-based Intelligent Driver Model (IDM) and Adaptive Cruise Control (ACC) in cooperative scenarios

). Rather than limiting themselves to on-board optimization at the level of an isolated vehicle, some researchers investigate the potential energy savings achievable through coordinated interaction among multiple vehicles. This approach rests on the idea that individual vehicle intelligence can be extended into collective intelligence, enabling dynamic coordination of speed profiles, maneuvers, and operating regimes across several vehicles simultaneously. The goal is to reduce overall energy losses by smoothing speed transitions, enhancing regenerative braking efficiency, and exploiting aerodynamic benefits.

Cooperative car-following for more economical driving (Eco-Car-Following):

In this approach, each following vehicle continuously adjusts its dynamics (speed, acceleration, torque) in response to the behavior of the preceding vehicle, aiming to minimize its own energy consumption while respecting safety (minimum distance) and comfort constraints. This is exemplified by the work in [85,171], which explicitly includes energy consumption in the objective function of their predictive controllers. Article [171] proposes an Economy-Oriented Car-Following Control (EOCFC) strategy based on MPC, where energy consumption is modeled as power demand. This formulation enables speed optimization while maintaining safe headway. Article [85] introduces an EACC (Energy-Optimal Adaptive Cruise Control) approach, also MPC-based, where motor energy is explicitly represented in the cost function. The key innovation lies in adding a predictive NARX model to anticipate the acceleration of the lead vehicle, thereby reducing instantaneous energy demand and extending battery life. These strategies promote smoother and more efficient driving by eliminating unnecessary speed fluctuations that lead to significant energy losses.

Cooperative platooning: group-level energy coordination:

Platooning approaches go a step further by considering the collective optimization of a group of connected vehicles. Here, researchers implement distributed predictive methods to jointly adjust the speeds, accelerations, and motor torques of all vehicles in the platoon. Article [172] proposes a CMOPC (Cooperative Multi-Objective Platoon Control) strategy based on multi-objective NMPC, combining longitudinal control (car-following) with torque allocation in vehicles with independent front/rear propulsion. Its originality lies in including the total energy consumption of the platoon in the cost function to achieve system-level optimization. Results show global energy savings of up to 4.7% depending on the driving cycle, highlighting the importance of coordinating powertrain control and platoon dynamics. In a complementary direction, article [173] focuses on maintaining string stability and constraint satisfaction within the platoon through a distributed reference governor (RG) mechanism. This RG intervenes only when predefined safety or comfort constraints are at risk of being violated, allowing the core controller to preserve string stability. Additionally, energy optimization is incorporated through a nonlinear model predictive controller (NMPC) dedicated to the plug-in hybrid leader vehicle, ensuring that fuel and electric power are managed efficiently. Hardware-in-the-loop (HIL) validation confirms that the method successfully balances constraint enforcement, stability, and energy efficiency across the platoon. Article [174] develops a DMPC method with an optimized regenerative braking strategy for following vehicles. This coordination not only improves braking efficiency but also maximizes energy recovery while respecting comfort and safety constraints. Some studies adopt a hierarchical structure combining collaborative planning at a higher level with detailed energy management at the operational level. For instance, article [175] proposes a two-layer approach: the upper layer performs cooperative speed planning using DMPC to ensure platoon stability, while the lower layer manages real-time energy distribution in a HESS (battery + supercapacitor) system. This reduces battery stress and leverages the dynamic capabilities of supercapacitors. Collectively, these studies show that cooperative platooning, when paired with distributed control strategies, enables optimization of the entire mobile system rather than each vehicle in isolation, offering significant gains in energy consumption and component lifespan.

Incorporating environmental and contextual information:

Adding environmental or contextual data enables further anticipation and optimization of driving behaviors based on upcoming road conditions. In this vein, article [176] combines terrain data (slope) with predictions of the lead vehicle’s motion via V2V communication. The approach is embedded in an NMPC framework, with a cost function that includes energy consumption at each control cycle. Dynamic optimization of motor torque and speed while accounting for the environment helps extend vehicle range. Article [86] targets complex urban scenarios, especially intersections, by combining offline speed planning via a Genetic Algorithm (GA) with SMC-based tracking control. By leveraging V2X data, the system anticipates intersection crossing to smooth speed transitions, resulting in up to 16% energy savings compared to rule-based speed planning in cooperative scenarios. Integrating road context and interactions with infrastructure or other vehicles significantly enriches the potential for energy optimization.

Despite these demonstrated benefits, several technological barriers still limit the practical deployment of these strategies:

• V2V communication reliability and latency: Urban environments often disrupt communication, undermining platoon stability and energy performance. Hybrid solutions or predictive controllers robust to communication loss warrant further exploration.
• Vehicle heterogeneity: Differences in state of charge (SoC), regeneration capacity, or thermal management strategies create imbalances. Adaptive models that account for the internal characteristics of each vehicle are necessary.
• High computational cost: The complexity of real-time predictive controllers remains a barrier to large-scale embedded deployment.
• Limited consideration of lateral behaviors (lane changes, merging): These dynamics, often overlooked, are essential for avoiding energy-wasting slowdowns. Spatio-temporal approaches using Signal Temporal Logic, for example, offer promising avenues.
• Underuse of regenerative braking as an explicit decision variable: Few strategies fully integrate regeneration potential with battery SoC considerations. A truly integrated co-optimization of energy recovery and driving tempo could greatly enhance overall efficiency.
• Aerodynamic vs. comfort trade-off: While driving at close distances within a platoon can significantly reduce aerodynamic drag, it also increases the likelihood of frequent accelerations and braking due to limited reaction time, ultimately leading to higher energy consumption. Future strategies must strike a balance between aerodynamic efficiency and energy comfort by dynamically adjusting inter-vehicle distances according to traffic conditions and vehicle capabilities.

3.2.6. Leveraging Vehicle-to-Grid (V2G) Capabilities for Onboard Energy Optimization

In parallel with these main axes, promising opportunities are emerging through interaction with the electrical grid, particularly via Vehicle-to-Grid (V2G) technologies. Although primarily developed to provide grid flexibility and stability [177,178,179], this bidirectional capability can also be leveraged to enhance onboard energy management in electric vehicles (EVs).

Underpinned by the ISO 15,118 standard [180], V2G enables the vehicle to act as an active, intelligent node within the energy system, able not only to draw power from the grid but also to deliver it back in a controlled, standardized, and secure manner. Some studies have explored onboard converter architectures and hierarchical control strategies specifically designed to facilitate efficient V2G/G2V energy exchange within the vehicle [181]. However, the direct implications of V2G for vehicle-range extension, onboard energy control, and battery health have received relatively limited attention in the literature.

Nonetheless, several mechanisms suggest that V2G could serve vehicle-level performance objectives beyond its contribution to the grid. For example:

• Dynamic charge adaptation: By considering planned trips, driving profiles, and terrain, the system could optimize the target state of charge (SoC), avoiding unnecessarily high SoC levels. Prolonged operation at high SoC is known to accelerate cell aging, which, over time, reduces the usable driving range [182].
• Controlled partial discharge: In certain scenarios, it may be advantageous to offload non-critical stored energy back to the grid to avoid maintaining a high SoC for extended periods—conditions that induce chemical stress and passive losses, especially at elevated temperatures [183].
• Intelligent thermal preconditioning: Taking advantage of low-tariff or off-peak hours, V2G systems can pre-heat or pre-cool the battery before departure, ensuring operation within its optimal thermal window. This not only improves immediate energy efficiency but also extends the thermal and chemical longevity of battery cells [184].

Although historically conceived as a grid-benefit technology, V2G holds clear potential to support contextual, vehicle-level energy optimization strategies in the future, with tangible benefits for range, battery durability, and overall operational efficiency.

4. Conclusions

Optimizing the energy consumption of electric vehicles (EVs) has become a central challenge to enable their large-scale deployment, meet expectations regarding range and sustainability, and contribute to the energy transition. The present work provides an original contribution by combining an in-depth bibliometric mapping with a targeted qualitative analysis of the recent literature, thereby offering a structured framework to understand research dynamics and identify the most relevant levers for action. In doing so, it addresses a crucial need: to provide a clear and hierarchical vision of the solutions explored within the scientific community in order to effectively guide future efforts toward improving EV energy efficiency.

Our quantitative analysis confirms a strong geographical concentration of publications, dominated by China. This leadership reflects the size of its domestic market, proactive public policies, and privileged access to large experimental datasets, thereby creating an ecosystem particularly favorable to innovation and explaining the prominent visibility of Chinese researchers in the literature on energy optimization for EVs. However, it is important to emphasize that other countries, notably in Europe, have also made significant contributions that complement and enrich the global understanding of the issue. German and French studies, in particular, stand out through methodological and experimental approaches that directly support the optimization of EV energy consumption: modeling and estimation works [185,186] improve demand prediction by combining detailed models with field data; comparative and experimental studies [187] provide valuable empirical validation; research on route planning and infrastructure integration [188] highlights the importance of routing to reduce overconsumption linked to stops; analyses of seasonal variability and auxiliary loads [189] show that heating and air conditioning can increase consumption by several tens of percent in winter; and finally, studies based on real fleet data and works integrating auxiliary management and automation [190,191] underline the need for robust models to ensure reliable applicability under operational conditions. These contributions represent essential complementary inputs, expanding the range of possible solutions beyond the Asian context alone. They remind us that research on EV energy efficiency is truly international and that its progress relies on the complementarity of diverse approaches and application contexts.

From a thematic standpoint, our corpus reveals a clear predominance of studies focusing on hybrid energy storage system (HESS) management and motor control/torque distribution. This orientation is technically and practically justified: HESS directly addresses the energy–power trade-off by protecting the battery from transient loads and improving energy recovery during braking, while torque control tackles the fundamental losses of the drivetrain and can often be implemented through software updates leveraging existing electronics. In contrast, levers such as thermal management or regenerative braking are less represented in the corpus—not because they are less relevant, but because they require complex multiphysics modeling, context-specific validations (climate and topography), and costly testing, which slows down their methodological diffusion. Systemic axes (V2V and V2G) appear promising but remain at an early stage, as they are constrained by dependencies on infrastructure, standardization, and regulatory frameworks.

These findings carry both operational and scientific implications: the focus on HESS and motor control reflects research oriented toward measurable, short-term, and testable gains, while less-addressed levers harbor high-impact opportunities in the medium term, provided that their methodological and experimental barriers can be overcome. This work highlights the need for more systematic integration across these axes—for instance, joint co-optimization of thermal management with HESS EMSs, or the explicit integration of battery aging into recovery strategies.

From a market perspective, the global momentum in EV adoption points to a significant increase in demand for lithium-ion batteries (LIBs) over the next decade. This expansion will have three major consequences for EMS research: (i) pressure on battery lifespan and total cost of ownership will reinforce the importance of management strategies aimed at limiting cell degradation; (ii) the diversification of fleets and usage profiles (commercial vehicles and shared fleets) will require more robust, adaptive EMSs capable of handling heterogeneous patterns of use; and (iii) the need to optimize the battery life cycle (including second-life applications and recycling) will impose a broadening of optimization criteria beyond instantaneous consumption (to include cost, carbon footprint, and battery health).

Consequently, the EMSs of tomorrow will need to be battery-aware and multi-objective: they must simultaneously optimize energy consumption, limit degradation, schedule charging according to price signals and grid conditions, and leverage V2G/V2H services where economically and technically relevant. This requires hybrid algorithms (predictive + learning-based) capable of integrating aging models, usage forecasts, and grid constraints, while remaining lightweight enough for embedded implementation.

Several research directions deserve to be further explored: (i) conducting systematic and standardized comparisons axis by axis on common benchmarks; (ii) developing experimental protocols and public datasets to ensure reproducibility and fair comparison of methods; (iii) carrying out multi-criteria studies integrating total cost of ownership and degradation to better inform industrial choices; and (iv) strengthening international coverage in bibliometric analyses to correct geographical biases and identify emerging practices in other regions. These efforts will complement and extend the contribution of the present study, which aims to provide both a mapping and a critical framework to orient future research toward truly integrated and industrially viable solutions.

In sum, this study demonstrates that reducing EV energy consumption does not rely on a single solution, but on the coordinated articulation of electrical, mechanical, and thermal subsystems. By proposing a critical and operational framework, it serves as a reference tool to guide future research toward more robust, integrated, and industrially deployable EMSs that are capable of supporting the rapid growth of the EV market and addressing the challenges arising from the increasing demand for lithium-ion batteries.